Limnol. Oceanogr., 55(1), 2010, 403–419 E 2010, by the American Society of Limnology and Oceanography, Inc. Two-step dialogue between the cladoceran Bosmina and invertebrate predators: Induction and natural selection

W. Charles Kerfoot1,* and A. Scott McNaught2

Department of Biology, University of Michigan, Ann Arbor, Michigan

Abstract Aquatic prey species respond to predators with fast (developmental) and slow (selective) feedbacks. Natural selection is assumed to fashion details of induction and to modify baseline morphology, but only rarely do we catch the slower (multi-generation) process in action. Laboratory experiments with Bosmina detected predator- mediated induction and estimated spine heritability (h2 5 0.2–0.5). Third Sister Lake, Michigan, U.S.A., Bosmina exhibited induction to resident (Mesocyclops) and to two nonresident, neighborhood predators (Epischura and ). However, the magnitude of induction in Third Sister Lake Bosmina to nonresident predators (Epischura, Leptodora) was muted, when compared with induction and final spine lengths in Epischura–Leptodora lakes. Inadvertent escape of Leptodora into Third Sister Lake in 1987 created a long-term (multiyear), whole-lake experiment, where resident Bosmina populations fell under intense size selection. During the interval, Leptodora suppressed a late-season smaller, short-featured species (B. freyii), favored seasonal expansion of an overwintering long-featured species (B. liederi), and selected for longer features in the latter species. Before local extinction, defensive spines of B. liederi achieved lengths comparable to populations that co-occur with Epischura and Leptodora.

Given the isolated and ephemeral nature of lakes, akin to between predators and prey (Tollrian and Harvell 1999; islands, to what degree are species and community Relyea 2002). From the standpoint of the prey species: (1) responses fashioned at the local (resident lake) or regional how often and quickly does the predator–prey inductive (lake district) level? Regional dispersal is receiving attention response evolve; (2) how is it maintained (physical–chemical, (De Meester et al. 2002; Havel and Shurin 2004), yet how developmental, and genetic feedbacks); and (3) how many does the duration of species contact modify predator–prey predators are involved? To answer these questions, we interactions, species persistence, and genetic adjustments require more information about the process of natural (Ricklefs and Schluter 1993; Tilman and Kareiva 1997)? selection in lakes, particulars of developmental and genetic Zooplankton communities are probably best approached responses, and details on neighborhood and historical from a regional metapopulation perspective (Kerfoot et al. ecological interactions. 2004; Leibold and Norberg 2004), because long-term In the mid–1980s, a series of zooplankton translocation survival of species is tied both to dispersal and to local experiments investigated predator–prey relationships at interactions (Shurin and Allen 2001). However, unraveling northeastern and Midwestern U.S. localities (i.e., moving ecological and evolutionary responses at local and regional predators and prey around in different lakes to explore scales requires extended studies, although insight may be food-web and ecological interactions [Kerfoot 1987; aided by fortuitous circumstances. McNaught 1993a,b]). Near the end of the planned manipu- One of the challenges is that predator–prey interactions lations in Third Sister Lake, Michigan, U.S.A., during the are complicated by two levels of response (developmental, summer of 1987, Leptodora inadvertently escaped from selective). Early on, species in the cladoceran genera Daphnia enclosures (McNaught et al. 2004). This highly efficient, and Bosmina were found capable of rapid developmental size-selective predator rapidly increased in abundance, reaction to the presence of predators (Daphnia: Grant and creating a long-term whole-lake experiment. Plankton Bayly 1981; Havel 1985; Bosmina: Kerfoot 1987). Since then, samples from 1987 to 1992 documented that Leptodora the nature of inducible defenses has been investigated from became abundant and differentially depleted small-bodied laboratory, field, and theoretical perspectives (Dodson cladoceran species (McNaught 1993a; McNaught et al. 2004). 1989a; Tollrian and Harvell 1999; Gabriel et al. 2005). Yet For 6 yr after 1992, as Leptodora reached the height of its abundance, Bosmina became locally extinct. When Leptodora there are many unanswered questions about the nature of crashed in the late 1990s, Bosmina reappeared in 1999 and the dialogue played out in ecological and evolutionary time became abundant again by 2004–2007. A sequence of translocation experiments, laboratory and field observations *Corresponding author: [email protected] between 1987 and 2007 allow insight into developmental and selective responses by Bosmina. Present addresses: 1 Lake Superior Ecosystem Research Center and Department of Biological Sciences, Michigan Technological University, Methods Houghton, Michigan 2 Department of Biology, Central Michigan University, Mount Study system and tested hypotheses—North of Ann Pleasant, Michigan Arbor, Michigan, there is a lake district that contains a 403 404 Kerfoot and McNaught

antennules protect the antennae, whereas the posterior mucrones protect rotation towards the ventral groove region, through which algae and bacteria are filtered. The early Third Sister Lake translocation experiments were designed to evaluate the relative importance of induction and natural selection in regulating defensive spine length during short-term experiments by challenging resident populations of Bosmina with more effective, nonresident predators. The unintentional escape and colonization of Third Sister Lake by Leptodora provided us an opportunity to see if natural selection over multiple generations by a highly efficient size-selective predator could push responses beyond regular induction norms. We divided the pheno- typic responses into three periods: pre-Leptodora, Lepto- dora, post-Leptodora. Fig. 1. Bosmina liederi from Third Sister Lake, illustrating position of general length (TL, total length; CL, carapace length) Seasonal pelagic sampling—Bosmina morphology and and defensive spine (ML, mucrones length; AL, antennule length) measurements. Mucro sutures and antennule segments are counts abundance was monitored by hauling a 0.3-m-diameter of cell boundaries that make up the terminal spine structures. plankton net (75 mm or 120 mm Nitex) vertically at a 16-m- deep, central station. Larger net tows used to quantify Leptodora abundance (Fig. 2) are discussed in McNaught diverse assemblage of invertebrate predators (McNaught et et al. (2004). All microcrustaceans were preserved in a 5% al. 2004). Two species of invertebrate predators (the formalin solution to which 40 g L21 of sucrose were added. calanoid copepod Epischura lacustris, the cladoceran Around 40 Bosmina were haphazardly removed from each Leptodora kindtii) not previously reported from Third plankton sample and mounted for measurement. Individ- Sister Lake were collected at Whitmore Lake for a series of uals were placed on a glass slide in a 50% glycerin–water experiments. Previous geographic surveys suggested that mixture. Slides were covered with a glass cover-slip and Third Sister Lake would be suitable for both species, measurements were taken under a Zeiss Universal Micro- although the lake was relatively small (McNaught 1993a,b; scope at 5003. Features measured on all individuals Kerfoot et al. 2000). The field experiments were done in included: (1) total body length, (2) number of mucro suspended enclosures (McNaught 1993a,b; McNaught et sutures, (3) length of mucro, (4) number of antennule al. 2004). Community interactions included size-selective segments, and (5) length of the antennule, measured from responses of planktonic communities to fish and inverte- the tooth to the tip (Fig. 1). If the antennule was curved, brate predators (McNaught 1993a,b; McNaught et al. measurements followed the curvature. 2004), whereas trait responses focused on Bosmina spine lengths (Fig. 1) relative to introduced Epischura and Heritability estimates for defensive traits—Reasons for Leptodora. estimating the heritability (h2) of defensive spines were two- Third Sister Lake was also selected because it had a fold. First, this information is important for determining history of limnological investigations. The waterbody is a the magnitude of selective responses, because the selective small (0.04 km2) seepage lake isolated somewhat to the response equals the heritability times the selective differen- south of the northern district as the third kettle-hole in a tial (Lynch and Walsh 1998). Secondly, we wanted to series that straddles Ann Arbor, Michigan (First, Second, emphasize the cost of selective responses (individuals lost) and Third Sister Lakes; McNaught et al. 2004). The lake is during selection relative to the neutral (no mortality) relatively deep (maximum and mean depths of 16.5 m and condition during induction. 7.2 m, respectively), with no permanent surface inlets or Heritability estimates were attempted only on B. liederi, outlets, and is moderately eutrophic (Lehman and Nau- the most abundant species. Electrophoretic and sequencing moski 1986). Water-column stratification occurs in early studies showed that Third Sister Lake has a cryptic two- spring, after a brief mixing period, followed by the species assemblage (Bosmina liederi and B. freyii), which hypolimnion quickly becoming anoxic (Bridgeman et al. complicated comparisons (W. C. Kerfoot and L. J. Weider 2000). By late summer the thermocline is at 5–6 m and unpubl.). Headpore placement was not useful in separating phytoplankton and zooplankton are concentrated in the the two species (Korinek et al. 1997; Taylor et al. 2002). upper 5 m. Phytoplankton assemblages are dominated by However, a morphometric character, spacing of teeth on diatoms, chrysophytes, and cryptomonads in the spring the postabdomenal claw, previously pointed out by De and autumn and by cyanobacteria in the summer. Melo and Hebert (1994), could be used to distinguish the Bosmina possess chitinous anterior (antennule) and two species in formalin-preserved samples. For cross- posterior (mucrone) protuberances (Fig. 1). When attacked comparisons, we also estimated trait heritability in B. by predatory copepods, they fold their single pair of liederi from Douglas Lake, Cheboygan County, Michigan, swimming appendages (antennae) into a groove behind the an Epischura–Leptodora lake. antennules and fall passively through the water column. Parthenogenetic cladocerans present favorable options During the motionless, dead-man behavior, the fixed for measuring heritability. The heritability (h2) of a trait is Bosmina and invertebrate predators 405

Fig. 2. As Leptodora became abundant, it depressed densities of small cladocerans, especially Ceriodaphnia and Bosmina. Leptodora’s appearance in Third Sister Lake is marked by the dashed vertical line. the fraction of the phenotypic variance that is attributable Laboratory induction experiments—Bosmina show devel- to genetic variation (Falconer 1960; Lynch and Walsh opmental induction when placed in close proximity to 1998). Under controlled laboratory conditions, interactions certain predators (Kerfoot 1987, 2006), although the between genotype and environment are assumed to be transmitting agent in different species is not always a negligible; hence, VP 5 VE + VG (VP 5 phenotypic chemical kairomone, but may also involve physical variance; VE 5 variance due to environmental influences; stimulation (Sakamoto et al. 2007). Laboratory induction VG 5 variance due to the different genotypes). We sampled experiments were carried out with Third Sister Lake lake populations and randomly transferred stem females to Bosmina and resident (rotifer Asplanchna sp., cyclopoid establish individual clones (i.e., a number of genotypes, Mesocyclops edax, Chaoborus punctipennis midge larvae) or grown together under common-garden conditions). Prior neighborhood (Epischura lacustris, Leptodora kindtii, Poly- to cloning, lake water was collected, millipore-filtered, phemus pediculus, Eurytemora affinis, Bythotrephes long- aerated, then stored for weeks to lower any residual imanus) invertebrate predators. The laboratory experiments kairomones. Bosmina were collected from the lake with a utilized a split clonal design. The paired design takes into 120-mm Nitex plankton net and individual females pipetted account close proximity effects on individuals (common into ,50 35-mL shell vials filled with millipore-filtered and environment effects on identical genotype) and demo- conditioned lake water. Cultures were fed daily with graphic composition, making the tests very sensitive Chlamydomonas sp., and medium exchanged weekly. (Kerfoot 2006). The exposure included both chemical and Clones were numbered and followed for three generations physical stimuli. Bosmina stem females were cloned within to purge maternal effects. Vial populations were preserved 35-mL glass shell vials on a medium that contained finely in a sucrose–formalin mixture and measured at 5003 under filtered (0.4-mm Millipore) and aged lake water. Chlamy- a Zeiss Universal. domonas reinhardii was added as food each day, medium Morphometric data were analyzed via a single classifi- was changed weekly, and new vials were started at about 2– 2 cation ANOVA (Sokal and Rohlf 1995). To estimate h B, 3-month intervals. After a minimum of three generations to the variance among the genotype means, the average purge maternal effects, around 2 weeks, a growing clone variance among clones was compared to the variance was split into two vials. One vial received a single predator within clones (Lynch and Walsh 1998). Within each clone, (e.g., advanced instar Mesocyclops edax, Epischura lacus- only one genotype is considered; hence, the observed tris, or juvenile Leptodora kindtii), whereas the other served variation is entirely environmental in origin. Heritability as a control. 2 in the broad sense is defined as: h B 5 VG/(VG + VE). Thus There was concern that large Leptodora could deplete the mean square within clones (MSwin) provides an estimate oxygen in small containers, so only immature instars were of VE. The mean square (MSbtw) between clones has utilized. The vial experiments were run for 7–12 d to test environmental as well as genotypic components, essentially for induction, after which the entire vial contents were VE + nVG (when n is the number of twins or clone-mates preserved in 10% formalin–glucose. We observed predators per vial; Parsons 1973, Lynch and Walsh 1998). So VG is in the vials, to check on condition. If predators died during estimated from (MSbtw2MSwin)/n. In addition, use of the experiment, they were replaced (only 21% of trials). If common-garden cultures during the interval that bridged there was obvious evidence for size-selective depletion of the Leptodora introduction allowed us to determine if field young during the brief exposure, the vial results were not populations were pushed beyond typical reaction norms. included in ANOVA tests (9% of tests). 406 Kerfoot and McNaught

Table 1. Heritability estimates for defensive spine traits in Bosmina leideri based on single-classification clonal ANOVA (Sokal and Rohlf 1995; Lynch and Walsh 1998). Shown are results from Third Sister Lake, Washtenaw County, Michigan, (1986) and two separate clonal rearing experiments from Douglas Lake, Emmet County, Michigan, U.S.A. (1989). In the table, degrees of freedom for between- clone mean squares (MS) is the number of clones 21. An example ANOVA analysis and heritability estimate is shown in 1B.

A. Heritability estimates Clones Individuals Heritability Fp Third Sister Lake Mucro length 8 47 0.38 4.5 8.8031024 Ant length 8 51 0.25 3.2 0.008 Douglas Lake—first test Mucro length 28 227 0.40 6.3 1.33310215 Ant length 28 188 0.33 4.4 1.8831029 Mucro sutures 28 198 0.44 6.5 2.95310215 Ant segments 28 192 0.43 6.1 4.52310214 Douglas Lake—second test Mucro length 22 112 0.53 6.8 3.63310211 Ant length 22 114 0.21 2.4 0.002 Mucro sutures 23 115 0.51 6.1 2.03310210 Ant segments 22 113 0.22 2.4 0.002 B. Example ANOVA (Douglas Lake, mucro length); MS5SS/df; SS5g(X2¯x)2 SS df MS Fp Between clones 14.9054 21 0.70978 6.8 3.63310211 Within clones 9.3852 90 0.10428 n55.1 VE50.10428 VG5(0.7097820.10428)/5.150.11873 Heritability5VG/VG+VE50.11873/(0.11873+0.10428)50.53

Field induction experiments—To test induction responses jars were returned to the laboratory, where Bosmina were under lake conditions, we used suspended transparent pipetted individually from the mixed-species field sample polyethylene bags. Populations of Bosmina respond differ- into a single jar with filtered Third Sister Lake water. After ently to size-selective Mesocyclops, Epischura, and Lepto- we returned to the lake, the Bosmina culture was stirred to dora because these taxa differ in body size and effectiveness mix individuals evenly throughout the vessel, and equal of prey removal (Kerfoot 1978; Hanazato and Yasuno portions were distributed to enclosure bags and collection 1989). Of the three, Leptodora is the most effective predator jars (Reference Samples). The reference samples provided on Bosmina, capable of depleting populations (Lunte and estimates of introduced numbers and initial morphology. Luecke 1990; Branstrator and Lehman 1991; McNaught et Epischura from Whitmore Lake were introduced into half al. 2004). Mesocyclops and Epischura were used primarily the bags at initial densities of 0.5 individuals L21, whereas in the mid-1980s because small-bag methods were perfected the other bags served as resident-predator removal (Kerfoot 1987), whereas a larger, deeper enclosure design treatments (see below). Bags were sampled periodically by for Leptodora was being tested (McNaught 1993a,b). The stirring and removing 10.5-liter subsamples to estimate smaller enclosures consisted of polyethylene bags (160 li- densities and Bosmina morphology. At the end of the ters; 50-mm wall thickness) suspended from floating experiment, all remaining water was filtered through a 75- platforms. The platforms were buoyed by styrofoam slabs, mm Nitex plankton net and organisms preserved in 10% arranged in rows, and anchored at their ends via cinder formalin and 40 g L21 sugar. blocks to the lake floor. Wooden cross-braces allowed We encountered difficulties measuring Bosmina induc- placement of four bags per floating platform, secured by tion responses in the larger Leptodora enclosure experi- pressure clamps. The bags draped down a meter into the ments. Leptodora proved such an efficient predator on water column. small-bodied cladocerans that only one Bosmina was Moored bag experiments with Mesocyclops removal and retrieved from predator treatments (McNaught 1993a;A. Epischura addition were run in 1985 and 1986. The bags Scott McNaught unpubl.). Hence for testing Leptodora were filled with 160 liters of coarsely filtered (75-mm Nitex) induction responses, we relied on laboratory tests. lake water pumped from mid-epilimnion depth. Once bags Differences between treatment responses in bag experi- were filled, water was not renewed, nor was there exchange ments were examined using a nested Analysis of Variance with lake water via a mesh boundary (i.e., closed bag). (Nested ANOVA; SYSTAT; Wilkinson 1989) on the two Using a vertical plankton tow, zooplankton were sampled treatments (Epischura addition, predator removal). The from Third Sister Lake and placed in 4-liter glass jars. The nested ANOVA design incorporated three levels: (1) Bosmina and invertebrate predators 407

Fig. 3. Reduction in defensive spine lengths when individuals are removed from lake waters and placed in isolated laboratory culture. Hollow symbols are field lengths, whereas solid symbols are lengths in laboratory cultures (after at least three generations, to purge maternal effects). Linear regressions are fit to the points and equations shown above scatter plots. Note that the regression slopes are quite low. Comparison done on transfers from August 1986. measurements within individual bags, (2) differences lineages, with 112–227 individuals for each estimate. An among replicate bags in treatments, and (3) evaluation of example ANOVA table is given in Table 1B, which lists the treatment effects. The nested ANOVA examined only trait mean square (MS) for between and within clones. Variation sizes at the end of each experiment. Mucro and antennule between clones was much greater than variation within a segment counts were excluded from the nested ANOVA single clone, because F-values were significant (p , 0.05) to analysis because these were discrete measurements. We highly significant (p , 0.01) for all comparisons. Herita- emphasize that controls served more as a predator-removal bility for mucro features (mucro length, mucro sutures) treatment, than a strict control, because all resident ranged between 0.38 and 0.53 (mean 5 0.45), whereas invertebrate predators were removed via filtration at the heritability for antennule features (antennule length, start of the experiment and only Bosmina were reintro- segments) was lower, ranging between 0.21 and 0.43 (mean duced. At the end of an experiment, individuals were 5 0.29; Table 1). Under directional selection, for traits with preserved in formalin–sucrose and enumerated as before. a mean h2 5 0.50, a two standard-deviation selective differential (the amount that the mean is moved during the Results generation of selection; Futuyma 1986; Lynch and Walsh 1998) would produce around a one standard-deviation Heritability of features—Third Sister Lake tests utilized selective response, and would cost the population about eight clonal lineages and 47–51 individuals. Douglas Lake 95% mortality. Consequently, rapid selection in enclosure tests were more extensive, using 22–28 separate clonal experiments and in the field should be accompanied by a 408 Kerfoot and McNaught

Fig. 4. Results of laboratory induction experiments. (Left panels) In Bosmina liederi from Third Sister Lake, there is modest elongation of spines in the presence of Mesocyclops (resident predator) and Epischura (nonresident predator). (Right panels) In B. liederi from Portage Lake (Epischura and Leptodora present), the elongation is much greater. Linear regression lines are fit to scattered points and equations listed above. See Tables 2–3 for more examples. major reduction in population density and an increased risk The matched comparison (Fig. 3) illustrates another of local population collapse. curious feature of defensive spines. Spines show atypical growth patterns and are proportionally largest in immature Laboratory induction experiments—When brought in individuals. Growth is not proportional to body length, from the field, Bosmina clones decreased mucrone and because there is only a slight positive or slight negative antennule spine lengths, a response seen before in New allometry from immature to mature instars (Y 5 Axk;k Hampshire field-lab transfers (Kerfoot 1987). Figure 3 ranges from 20.32 to +0.68 for mucrones and 20.25 to compares August 1986 field populations of B. liederi with +0.18 for antennules). mixed laboratory cultures drawn from the lake during Spine lengths did not respond to several invertebrate summer and held in culture for over three generations to predators (Asplanchna, , Eurytemora, Chao- purge maternal effects. In the lake, mean mucro lengths borus, Bythotrephes) that reside in Third Sister Lake (TSL) ranged between 35 mm and 41 mm and mean antennule or in neighborhood regions. However, when placed in the lengths 5 108–119 mm, whereas after three generations in presence of resident Mesocyclops or neighborhood the laboratory in the absence of predators, mucro lengths Epischura and Leptodora, defensive spines (mucrones and of clones ranged between 20 mm and 25 mm and antennule antennules) showed significant elongation (Fig. 4; Table 2). lengths between 91 mm and 102 mm. The observed regres- Within 7–8 d of exposure, mucrones increased from 19– sion amounted to a 45% reduction in mucro length (2.4 28 mm to 29–35 mm(n 5 9 split clone experiments, mean standard deviations [SD]) and a 17% reduction in difference 695% C.L. 5 6.1 6 1.6 mm), whereas antennules antennule length (1.5 SD), although the absolute reduction increased from 88–101 mm to 102–122 mm(n 5 9, mean (,15–20 mm) was similar for both features. Differences difference 5 13.6 6 3.4 mm). Paired ANOVAs confirmed between field and lab were highly significant (two-sample t- highly significant increases between predator treatments test with unequal variance, mean 6 95% confidence limits and controls in almost all individual experiments (Table 2). (C.L.): mean field mucro length, 39 6 1 mm; lab 23 6 1 mm, Mucro length increased 1.0–3.1 SD (mean 5 1.5 6 0.5 SD) t 5 21.5, df 5 187, p 5 8.6 3 10253; mean field antennule and antennule length 1.2–3.8 SD (mean 5 1.9 6 0.6 SD). length 5 115 6 2 mm, lab 99 6 2 mm, t 5 12.0, df 5 192, p Yet how do the Third Sister Lake responses compare 5 2.5 3 10225). with Bosmina responses in lakes where the three taxa Bosmina and invertebrate predators 409

Table 2. Laboratory induction experiments (split-clonal in 35-mL shell vials) from Third Sister Lake, Michigan, U.S.A. Results listed by predator, source of clone and date, trait (mucro, antennule length), number of individuals measured (n) and treatment (predator, control), mean length (Mean) and standard deviation (SD) of trait. ANOVA compares controls vs. predator treatment in split-clone tests; F-values, degrees of freedom (df), and p-values are given after trait. Predators are Mesocyclops edax, Epischura lacustris, and Leptodora kindti. Clone TSL10 is B. freyii, whereas TSL4-6 clones are B. liederi.

Predator Clone Variable n Treatment Mean Fp-value Mesocyclops TSL4-05 Mucro 44 Control 28(4.8) 38.7 1.9031028 41 Predator 34(3.6) (1,83) Antennule 44 Control 101(9.4) 112.8 3.88310217 41 Predator 122(8.5) (1,83) Mesocyclops TSL6-05 Mucro 46 Control 26(4.6) 17.8 5.9331025 43 Predator 30(3.7) (1,87) Antennule 46 Control 100(7.1) 27.9 9.3131025 43 Predator 109(8.3) (1,87) Epischura TSL5-05 Mucro 42 Control 27(5.3) 17.7 8.0031025 25 Predator 32(4.4) (1,65) Antennule 42 Control 99(7.9) 30 7.5531027 25 Predator 111(11.1) (1,65) Epischura TSL6-05 Mucro 32 Control 24(2.9) 120.4 1.11310214 18 Predator 35(4.2) (1.48) Antennule 32 Control 98(8.6) 21.3 2.6331025 21 Predator 108(4.2) (1,51) Epischura TSL10-05 Mucro 42 Control 26(5.7) 17.7 8.0031025 25 Predator 32(4.3) (1.65) Antennule 42 Control 99(8.2) 30.0 7.5531027 25 Predator 112(11.0) (1,65) Epischura TSL6-06 Mucro 35 Control 26(4.2) 37.2 8.3231028 27 Predator 32(3.6) (1,60) Antennule 35 Control 93(8.5) 47.5 3.8431029 27 Predator 107(6.4) (1,60) Epischura TSL6-06 Mucro 35 Control 23(3.4) 38.5 3.7231028 35 Predator 30(4.8) (1,68) Antennule 35 Control 95(7.1) 86.0 1.10310213 35 Predator 110(6.2) (1,68) Epischura TSL10-06 Mucro 34 Control 23(4.2) 37.2 5.7131028 36 Predator 29(4.1) (1,68) Antennule 34 Control 91(5.8) 118.7 1.47310216 36 Predator 110(8.9) (1,68) Leptodora TSL5-05 Mucro 34 Control 23(4.0) 35.8 1.1731027 30 Predator 29(3.4) (1,62) Antennule 34 Control 88(9.9) 40.3 2.8631028 30 Predator 102(8.0) (1,62) Leptodora TSL10-06 Mucro 36 Control 24(4.0) 8.8 3.8831023 49 Predator 27(3.2) (1,84) Antennule 36 Control 97(9.0) 0.4 0.5431021 49 Predator 96(15.8) (1,84) Leptodora TSL5-06 Mucro 12 Control 23(3.2) 18.7 1.2731024 24 Predator 31(5.3) (1,35) Antennule 12 Control 88(8.3) 5.6 2.3931022 24 Predator 100(15.9) (1,35)

(Bosmina, Epischura, Leptodora) naturally co-occur? In- 3.8 mm; antennule 21.0 6 8.6 mm). Comparisons of the duction experiments using B. liederi and Epischura from mean magnitude of induction (Third Sister Lake vs. Portage Lake, Houghton County, Michigan, where all Portage Lake) demonstrated significantly greater responses three taxa co-exist, show two pronounced differences. First, in Bosmina in an Epischura–Leptodora lake (ANOVA, 26 regressed mucro lengths were shorter in Third Sister clones n 5 11; mucro responses F1,10 5 85.0, p 5 3.3 3 10 ; than in Portage Lake clones (Third Sister, 20–25 mm; antennule F1,10 5 7.0, p 5 0.02). In terms of SDs, mean Portage, 37 mm). Second, induction in TSL clones in- mucro length in Portage Lake induction tests increased 3.1 creased spine length only 40–60% the amount of induction SD, whereas mean antennule length increased 2.3 SD. observed in Portage Lake clones (Figs. 3,4; Table 2; n 5 3, What if one predator is entirely absent from the regional Portage Lake mucro mean difference 695% C.L. 5 18.3 6 neighborhood? For this comparison, split-clone induction 410 Kerfoot and McNaught

Table 3. Induction responses of Bosmina from an Epischura–Leptodora lake (Bosmina liederi; Portage Lake, Houghton County, Michigan, U.S.A.) and from Europe (; Epischura regionally absent). Epischura added to 35-mL vial for 7–10-d exposure (clone number and year of test). Variables include: sample size (n), treatment (predator, control), mean trait length (SD).*

Region and clone Variable n Treatment Mean Fp-value North American PLA-05 Mucro 42 Control 37(5.4) 163.4 3.25310221 42 Epischura 55(6.8) (1,82) Antennule 42 Control 93(9.9) 86.7 1.71310214 42 Epischura 112(8.9) (1,82) PL1-05 Mucro 72 Control 37(5.2) 436.3 8.67310247 84 Epischura 54(5.2) (1,154) Antennule 72 Control 90(8.4) 345.5 3.46310241 84 Epischura 115(8.4) (1,154) PL2-05 Mucro 42 Control 37(5.4) 168.3 1.15310221 43 Epischura 55(6.7) (1,83) Antennule 42 Control 93(9.9) 89.8 7.33310215 43 Epischura 112(8.8) (1,83) European GPS1-06 Mucro 43 Control 15(3.9) 0.3 0.58 Grosser Ploener See 45 Epischura 15(4.9) (1,86) Antennule 43 Control 75(11.0) 1 0.33 45 Epischura 77(8.9) (1,86) GPS54-07 Mucro 42 Control 16(3.2) 6.3 0.01 Grosser Ploener See 41 Epischura 18(4.2) (1,81) Antennule 42 Control 80(8.7) 11.7 9.9731024 41 Epischura 86(7.6) (1,81) SEL2-06 Mucro 42 Control 22(6.2) 0.2 0.65 Selenter See 42 Epischura 21(4.9) (1,82) Antennule 42 Control 90(13.7) 1.4 0.24 42 Epischura 94(11.6) (1,82) * ANOVA results on control vs. predator treatment in split-clone experiments includes F-value (df), and probability (p). The Epischura came from the same source (Portage Lake). experiments were carried out with B. longirostris from (Reference Samples). By the end of the 9-d exposure, mean German lakes, where Epischura does not occur, and lengths reduced to 20–23 mm in removal treatments (25% Portage Lake Epischura lacustris. When placed in labora- reduction), yet increased to 35–46 mm in predator-addition tory culture without predators, the small European treatments (29% increase). Antennule lengths began at 68– bosminid regressed to the shortest spine lengths of all 112 mm (Reference Samples). Mean antennule lengths Bosmina tested (15–22 mm mucro; 75–90 mm antennule). In declined to 84–90 mm in removal treatments (28% reduc- the presence of Epischura, two of three clones showed no tion), yet were maintained at 102–116 mminEpischura significant elongation of mucrones. Mean antennule addition treatments. A nested ANOVA (Table 4) run on elongation was only 4 mm, the lowest response in Bosmina the midsummer end samples (19 Jul) showed significant tests, and was only significant in one of three tests treatment effects on body length (F1,6 5 3.6, p 5 0.002), (Table 3). In terms of SDs, the mean response for mucro mucro length (F1,6 5 4.7, p , 0.000) and antennule length length was 0.1 SD and 0.4 SD for antennule length (i.e., (F1,6 5 5.6, p , 0.000). hardly any induction at all). An identical enclosure experiment run later in the year (02–19 Aug) gave very similar results (Table 4). Although Field enclosure experiments—Induction was also tested in Bosmina density was greater in late-season bags and the field enclosure experiments (Tables 4–5) with predator- duration of predator exposure longer, the outcome was removal and Epischura treatments. Epischura were introduced nearly identical. Defensive spine lengths (mucrones and at a density of 0.5 L21 and were present in all predation bags antennules) were significantly shorter in predator-removal at the end of the experiment. Bosmina survived the duration of treatments and longer in Epischura addition treatments 242 the experiment as 10.5-liter subsamples on 19 July 1985, (ANOVA, mucrones F1,4 5 76.7, p 5 4.9 3 10 ; 228 retrieved a mean of 41(SD 5 9.9) Bosmina from controls and antennules F1,4 5 44.3, p 5 6.2 3 10 ). Mean mucro 31(SD 5 27.8) from predation treatments. lengths were 22–23 mm in removal treatments and 36– Figure 5 illustrates trait responses in bags with addition 40 mminEpischura addition treatments. Mean antennule of Epischura lacustris, relative to predator-removal treat- lengths were 88–93 mm in predator-removal treatments and ments. Mean mucro lengths began at around 15–29 mm 105–109 mm in predator-addition treatments. Bosmina and invertebrate predators 411

Table 4. Results of in situ enclosure (plastic bag) experiments, Third Sister Lake, summer 1985. Reference sample represents a subsample of the Bosmina placed into enclosure bags. Only results from end sample dates are shown. Controls are actually a Mesocyclops- (predator) removal treatment, where there was no water exchange between bag and lake. All lengths (L) are in micrometers. Trait values are means 6 95% C.L. (SD) with ranges underneath (n5sample size).

Date Sample n Total L Mucro L Antennule L First enclosure experiment (11–19 Jul 1985) 11 Jul Reference 50 353620(71) 2861(4) 11663(10) 224–465 15–29 68–112 19 Jul Control 2 47 310619(63) 2061(4) 9063(8) 224–465 15–29 68–112 19 Jul Control 3 38 309614(43) 2261(4) 8463(9) 232–403 14–32 67–101 19 Jul Control 5 28 337619(48) 2261(3) 8463(7) 226–438 14–32 72–99 19 Jul Control 6 50 287617(59) 2361(4) 8563(10) 203–429 14–32 64–104 19 Jul Epischura 1 13 318643(71) 3565(8) 10369(14) 215–476 23–49 75–121 19 Jul Epischura 4 23 340631(71) 3765(12) 10267(14) 254–492 13–51 78–126 19 Jul Epischura 7 15 328653(96) 4664(7) 11668(13) 238–483 32–55 97–132 19 Jul Epischura 8 72 296616(70) 3962(7) 11062(9) 216–520 18–53 92–130 Second enclosure experiment (02–19 Aug 1985) 02 Aug Reference 50 332617(59) 2361(4) 10563(11) 249–551 15–31 77–124 14 Aug Control 2 50 302618(62) 2361(5) 9363(10) 205–434 12–33 75–113 14 Aug Control 4 52 28569(32) 2262(6) 8863(9) 217–383 14–55 72–122 19 Aug Epischura 1 50 339617(61) 3862(6) 10963(10) 223–447 26–53 86–128 14 Aug Epischura 3 45 327618(60) 4063(10) 10564(12) 235–495 14–55 78–128 14 Aug Epischura 8 52 301617(61) 3662(7) 10562(8) 209–444 17–49 84–122

These experiments confirm that the timing and magni- that morphological variance between individuals was less in tude of field responses corresponded closely to laboratory predator-removal treatments (controls) than in predator- induction experiments, resulting in 2–3 SD differences addition treatments (Table 5; Fig. 5). Note that the between predator-removal and predation addition treat- variance observed between individuals in predator-addition ments within 7–17 d, or about 1–2 generations. The treatment bags also compared favorably with the variance morphological responses were reversible, increasing when observed in nature (e.g., in Reference or routine field predators were added and reducing when they were samples). Hence, in the absence of a localized stimulus from removed. This correspondence argues that field responses a predator, clones regressed back towards a more uniform were due largely to short-term developmental (phenotypic) morphology. induction, rather than to directional selection. No major population reduction was observed in field Seasonal sampling of Third Sister Lake—Pre-Leptodora enclosures, as Bosmina density decreased only slightly in period: Total length tended to be longer in late winter and July predation treatments, from 0.8 (SD 5 0.4) to 0.5 (0.2) spring (371–445 mm), and shorter in late summer (298– individuals L21, whereas control densities were increasing 401 mm). Some of the seasonal trends were due to from 0.9 (0.4) to 3.9 (0.5) individuals L21. In August demographics, because early spring populations were enclosures, Bosmina densities increased in both control dominated by overwintering females of B. liederi.Mor- (from 4.1 [1.2] to 32.8 [38.1] individuals L21) and in phological patterns for 1984–1985 (before Leptodora predation treatments (from 4.3 [2.1] to 45.6 [46.3] introduction) indicated that Bosmina spines were longer individuals L21). in early spring (Fig. 6). Both features began a decline in Morphological variance patterns among bags provided June that continued through July and early August. another insight (Gabriel 1999; Gabriel et al. 2005). The Detailed inspection of seasonal dynamics disclosed that nested ANOVA at the end of the field experiment revealed the resident population was composed of two very cryptic, 412 Kerfoot and McNaught

Table 5. Nested ANOVA run for end samples (19 Jul 1985) of first enclosure experiment, Third Sister Lake. Treatments were addition of Epischura lacustris (Invertebrate Predator, initial density 5 0.5 L21) and Control (predator removal) in suspended plastic bags. Hypothesis is a test of significance between control and predation treatments, whereas control and Epischura are tests for significant differences between bags in that category. Traits measured included total body length, mucro length, and antennule length. Note that, at the end of the experiment, mucro and antennule lengths in control bags are much more uniform than in predator treatments.

Source Sum of squares df Mean-sq Fp Total body length Hypothesis 87,233 6 14,539 3.6 0.002 Control 46,162 3 15,387 3.8 0.011 Epischura 41,071 3 13,690 3.3 0.020 Error 1,139,633 278 4099 Mucro length Hypothesis 1084 6 181 4.7 0.000 Control 189 3 63 1.6 0.181 Epischura 894 3 298 7.7 0.000 Error 10,670 277 39 Antennule length Hypothesis 3192 6 532 5.6 0.000 Control 1024 3 341 3.6 0.014 Epischura 2168 3 723 7.6 0.000 Error 24,678 260 95 but morphologically and electrophoretically distinguish- (Fig. 5; McNaught et al. 2004), Bosmina became seasonally able species (W. C. Kerfoot and L. J. Weider unpubl.). truncated compared to pre-Leptodora years. Late-season Differentiating the species helped clarify some of the truncation could have contributed to the observed population observed seasonal trends. Both species were present in collapse in 1992, because Bosmina typically produce over- early spring samples, but one was much more abundant (B. wintering eggs in late September to November. liederi). The relative abundance of B. freyii increased in late When viewed over a sequence of years (1983–1991), summer and autumn, moving from 2% to 11% in April– summer samples document increased mucro and antennule May to 55% to 93% by August–September (Table 6). Early lengths following Leptodora introduction (Fig. 7). Popula- season heterogeneity was evident in the relative variance of tions had relatively short mucrones and antennules in 1983, features measured during April and May. Prior to 1987, and variable moderate lengths between 1984 and 1987. Coefficients of Variation (SD divided by the mean) for After introduction of Leptodora (arrow), mucro lengths traits were high throughout the season (mucro length, 17.0– increased up to around 45–52 mm and antennule lengths to 27.2%; antennule length, 9.1–14.0%), partly reflecting the 150–158 mm, similar to lakes where Bosmina co-occur with mixture of species. Spine lengths in both species were Epischura and Leptodora (Table 2), remaining long until greatest in early spring and declined over the course of the Bosmina population collapse in 1992. summer, although B. freyii had relatively shorter mucrones Evidence for long-term selective responses came from and antennule lengths. laboratory common-garden culture comparisons. As part of induction and heritability studies, Bosmina were brought into Leptodora period: Inadvertent introduction of Leptodora the laboratory and cultured periodically, raised under nearly in 1987 resulted in dramatic declines in small-bodied identical rearing procedures. Comparison of 15 common- cladocerans (Fig. 2; Bosmina, Ceriodaphnia). After Leptodora garden cultures from 1986 with 26 from 1989 and 3 from introduction, the seasonal pattern of Bosmina traits also 1991 indicated significant differences between pre-Leptodora changed (Fig. 6). Seasonal samples from 1990 to 1991 showed and Leptodora years. Some individual culture morphologies that Bosmina populations maintained long features through- are given in Table 7, whereas 1986 vs. 1991 clones are out the year, with no indication of a late-season decline. contrasted in Fig. 8. Comparing 1986 with 1989 lab cultures, Lengths of both traits (defensive spines) also increased. Mean laboratory-acclimated cultures exhibited highly significant mucro lengths ranged between 45 mmand52mm, whereas increases in mucro and antennule lengths (t-test 1986 vs. antennule lengths ranged between 150 mmand158mm. The 1989; mucro df 5 120, t 5 21.1, p 5 1.58 3 10242;antennule smaller featured species (B. freyii) was severely depressed or df 5 93, t 5 14.9; p 5 1.22 3 10226). Increases in body length not present in late season samples (Table 6). The observed were additionally seen, although the responses were less decline of coefficients of variation (mucro length, 10.0–17.2%; impressive than observed for spine lengths. Using pair-wise antennule length 5.1–8.9%)inLeptodora samples also comparisons with the difference expressed in SD units, the suggested a single species (i.e., B. liederi). Because Leptodora amount of morphological shift between pre-Leptodora (1986) reached maximum abundance in August to early September and Leptodora period (1989–1991) in common-garden Bosmina and invertebrate predators 413

Fig. 5. Results of induction experiments in field enclosures. The right two panels show responses of spines in removal and addition experiments. In control bags where all predators were removed, spine lengths decreased, whereas in bags where Epischura were added, spines remained long or even increased in length. Initial lengths of spines in the treatments are shown in the left insert panels. Plotted symbols are means, 6SD, and ranges (hollow symbols). Table 4 gives additional measurements. cultures was 1.7–3.7 (mean 5 2.8) SDs for body length, 2.2– 1992. The mean size of mucrones and antennules in 1989– 5.7 (mean 5 3.8) SDs for mucro length, and 1.5–4.8 (mean 5 1991 cultures corresponded closely to B. liederi populations 3.3) SDs for antennule length. Responses of about 1.5–2.0 that naturally co-exist with Epischura and Leptodora SDs in field populations for spine lengths were expected from (Fig. 4b; Table 3). induction alone (Table 2). However, the common-garden Post-Leptodora period. After Leptodora crashed, Bos- mean shifts indicate directional selection also underlying mina appeared again in 1999. Both species (B. liederi, B. spine length shifts of B. liederi prior to population collapse in freyii) reappeared, with B. liederi most abundant in spring 414 Kerfoot and McNaught

Fig. 6. Seasonal pattern for Bosmina spine lengths (mucrones, antennules) in Third Sister Lake, before and after Leptodora colonization. Left and right figures present composite diagrams for seasonal samples 2 yr before (1984–1985) and after (1990–1991) Leptodora introduction. Box plots give mean (square), median, 25th, and 75th percentiles; whiskers range from 10th percentile (bottom) to 90th percentile (top).

Table 6. Frequency of Bosmina freyii and B. liederi in Third Sister Lake samples before (1984–1985), during (1989–1991), and after (1999–2004) Leptodora colonization. NR 5 no record.

1984–1985 1989–1991 1999–2004 Month n B. freyii B. liederi n B. freyii B. liederi n B. freyii B. liederi Apr 70 11% 89% 42 0% 100% 59 3% 97% May 67 2% 99% 101 1% 99% 60 22% 78% Jun 148 16% 84% 42 0% 100% NR NR NR Jul 84 5% 94% 43 2% 98% 70 21% 79% Aug 40 55% 45% 41 0% 100% 119 28% 61% Sep 28 93% 7% 98 1% 99% NR NR NR Bosmina and invertebrate predators 415

details of spine development. The observed atypical growth (bradytelic pattern) of defensive spines produces lengths that are proportionately largest in the first few instars. This aberrant developmental growth pattern is shared among all tested species within the subgenus Bosmina (B. liederi, B. freyii, B. longirostris). Such a growth pattern emphasizes greatest risk in the smallest instars, and reduced risk in large, mature individuals. The size-dependent risk pattern of Bosmina fits the observed size-selective tactics of predatory copepods (Mesocyclops, Epischura; Kerfoot 1978; Chang and Hanazato 2003) and Leptodora (Chang and Hanazato 2004). Strong size-dependent morphological patterns are also found in Chaoborus predation on Daphnia (Tollrian 1995). Short-term laboratory experiments demonstrate that Bosmina in Third Sister Lake respond to resident Mesocy- clops edax by developmentally increasing defensive spine lengths within days. Induction experiments with neighbor- hood predators indicated that Third Sister Bosmina also respond to larger bodied nonresident predators (Epischura lacustris, Leptodora kindtii), present in nearby lake districts. Our induction experiments closely resembled earlier (1980– 1986) induction results in Vermont and New Hampshire (U.S.A.) lakes, where Bosmina transferred from a small pond to an Epischura lake responded to this predator by rapidly increasing spine lengths (Kerfoot 1987). One possible explanation for neighborhood sensitivity is that responses reflect a historical legacy fueled by regular regional encounters. Over the northeastern United States and Canada, Bosmina co-occur with Mesocyclops edax, Epischura lacustris, and Leptodora kindtii in 72%,65%, and 45% of regional lakes (Carter et al. 1980). Co-occurrence with Leptodora is more restricted because this species generally inhabits deeper lakes (Carter et al. 1980; McNaught 1993b). Induction was not stimulated by several other regional predators (Asplanchna, Chaoborus, Polyphemus), nor by two nonindigenous genera (Eurytemora, Bythotrephes). Although a temorid of similar body size to Epischura, Eurytemora affinis is a poor predator on Bosmina,isan introduced species, and occurs in only 0.1% of lakes (Carter et al. 1980). Polyphemus occurs in only 13% of lakes and also is not a particularly efficient predator on Bosmina (Carter et al. 1980). Asplanchna and Chaoborus are much Fig. 7. Yearly mid- to late-seasonal spine (antennule, mu- more efficient predators on rotifers and Daphnia, respec- crones) responses before, during, and after Leptodora appearance. tively, than on Bosmina. Bosmina longirostris from Europe, The times of Leptodora presence are indicated by arrows. Symbols which does not co-occur with any known Epischura species, as in Fig. 6. showed a very weak induction response to Epischura, underscoring that prolonged lack of contact can result in and B. freyii present in late summer to autumn (Table 6). loss of the developmental feedback. Mucro and antennule lengths varied, but reverted to pre- Previous regional comparisons of Bosmina demonstrated Leptodora patterns (Fig. 7). a strong geographic correlation between morphological features (antennule, mucro, and body size lengths) and risk Discussion related to invertebrate predators (Kerfoot 1980, 1987; Sprules et al. 1984). Modification of predators within Third In terms of a dialogue, prey sense the presence of danger Sister Lake enclosures showed that rapid developmental via kairomones or physical cues, and modify behavior or responses (induction) dominated short-term interactions. development within minutes to days to reduce individual Features rapidly increased in length in the presence of large risk (Dodson 1989b; Tollrian and Harvell 1999). However, predatory invertebrates, and would diminish down to small over longer spans (multi-generation), selection tailors sizes when all predators were removed. The shortening of 416 Kerfoot and McNaught

Table 7. Comparison of spine features in 1986, 1989, and 1991 common garden cultures (laboratory regressed) from Third Sister Lake. Measurements shown below are examples from 26 (1986), 25 (1989), and 4 (1991) preserved cultures. Lengths are in microns with mean695% C.L. (SD) and range below.

Date n Total length Mucro length Mucro sutures Antennule segments Antennule length 1986 No. 6 13 285629(49) 2162(4) 0.260.3(0.4) 9.860.7(1.1) 9664(7) 242–390 16–29 0–1 8–12 79–106 1986 No. 8 14 287631(54) 2463(5) 0.560.4(0.7) 9.560.5(0.9) 9264(8) 227–392 18–33 0–2 8–11 79–102 1986 No. 9 21 277622(48) 2161(3) 0.160.2(0.4) 8.960.4(0.8) 9664(9) 203–393 16–26 0–1 7–10 75–121 1986 No. 11 15 280630(53) 2362(3) 0.260.2(0.4) 9.560.6(1.1) 9965(5) 227–384 16–27 0–1 8–12 81–115 1989 No. 1 13 271615(26) 3863(5) 2.560.4(0.7) 11.860.5(0.9) 13065(8) 220–313 32–52 2–4 11–14 113–142 1989 No. 2 20 27967(16) 3762(4) 2.660.2(0.5) 12.460.5(1.1) 12766(13) 244–319 32–46 2–3 11–15 90–142 1989 No. 3 8 292640(48) 4164(4) 2.560.4(0.5) 12.861.2(1.5) 128611(13) 252–371 35–46 2–3 10–15 104–145 1989 No. 4 9 260627(35) 3664(6) 2.760.7(0.9) 12.261.4(1.4) 117615(19) 215–325 26–44 1–4 9–14 93–145 1989 No. 5 15 269617(30) 3762(4) 2.260.4(0.7) 11.960.8(1.5) 12266(11) 226–336 29–44 1–3 9–14 102–142 1991 No. 1 13 425627(45) 4163(6) 2.360.4(0.6) 12.560.6(1.1) 12667(12) 351–516 29–49 1–3 11–14 107–148 1991 No. 2(B) 24 392630(71) 4464(9) 2.860.4(0.9) 12.960.4(1.1) 13565(13) 287–534 20–55 1–4 11–15 99–154 1991 No. 3(A) 9 447634(44) 4265(7) 2.460.6(0.7) 13.060.9(1.1) 12967(9) 371–513 32–49 1–3 11–14 113–139 features observed in field enclosures (predator-removal After Leptodora crashed, Ceriodaphnia and Bosmina are treatment) closely resembled the reversible induction beginning to return to prior abundance. The Bosmina responses seen in laboratory culture, which occurred when community is reverting to earlier patterns, both in species prey were separated from predators. composition (B. liederi, B. freyii) and seasonal incidence. However, Third Sister Lake clones paired with Epischura Rapid return may have been aided by dormant eggs that in short-term trials initially did not achieve spine lengths bridged the decade interval (sediment core studies; W. C. observed in Epischura and Leptodora lakes, suggesting that Kerfoot unpubl.). natural selection also modifies lengths over geographic Reductions in bosminid spine lengths when individuals regions, finely tuning responses to resident predators. We are shifted from the field to the laboratory are usually suspect that reverse selection from competition may also interpreted as evidence for metabolic or demographic (e.g., relax the response. Both feature lengths (mucro, antennule) molting) costs (Kerfoot 1977, 1987, 2006). Early studies on showed moderate heritability (h2 5 0.2–0.5) in the Meso- Bosmina provided evidence for reduced clutch size and cyclops lake (Third Sister) and in an Epischura–Leptodora population growth associated with longer featured species lake (Douglas Lake), indicating potential for selective relative to shorter featured species (Kerfoot 1977; Kerfoot responses. Third Sister Bosmina could maintain population and Pastorok 1978). Recent work with Eubosmina demon- density in the presence of Mesocyclops and moderate strated increased drag associated with very long antennules concentrations of introduced Epischura, illustrating the or dorsal humps (Lagergren et al. 1997; Lord et al. 2006). importance of induced spine lengths reducing mortality and In long-term laboratory selection experiments with protecting individuals. However, natural selection could Epischura, morphological divergence of Bosmina lineages not be tested effectively in enclosure experiments, for occurs under predation and control treatments, because exposures were short, no more than 1–2 generations (7– predation selects for longer spines at the same time as 14 d). The inadvertent escape and colonization of Third competition in controls selects for diminished length (W. C. Sister by Leptodora afforded an opportunity to look for Kerfoot unpubl.). Loss of predator-resistant characteristics selection on Bosmina during an episode of enhanced size- is commonplace for lineages of algae reared for extended selectivity. After establishment, the density of the inverte- periods of time in the laboratory, presumably because brate predator increased, producing a 5-yr size-selective competition favors reduced anti-consumption features (W. episode that coincided with severe depression of small R. DeMott unpubl.). cladocerans (Bosmina and Ceriodaphnia). Common garden A widely discussed prerequisite for the evolution of cultures verified genetic shifts in B. liederi, pushing means induction is temporal variability (Gabriel 1999; Tollrian to levels commonly observed in Epischura and Leptodora and Harvell 1999; Gabriel et al. 2005). Several levels of lakes. temporal variability were evident in Third Sister Lake Bosmina and invertebrate predators 417

1991; Chang and Hanazato 2003). Additional excellent long-term examples of rapid evolutionary responses from sediment core studies include Cousyn et al. (2001) and Kerfoot and Weider (2004), where both studies showed that Daphnia develop anti-predator traits in the presence of predators and then lose them when predation is relaxed. Thus the predator–prey dialogue involves a restricted number of invertebrate predators, is spatially and tempo- rally dynamic on a regional scale, and seems fine-tuned by resident predators on a local scale. Bosmina liederi is the largest of the North American small-bodied species in the subgenus Bosmina, and often co-occurs with Leptodora and Epischura (Taylor et al. 2002). The observation that high concentrations of Leptodora may eliminate B. freyii and severely suppress Bosmina liederi, suggests that long-term coexistence be- tween this predator and the two prey species would be enhanced by even larger body size and longer spines. Evidence for a similar circumstance is evident in Lake Suwa, Japan, enclosure experiments, where Leptodora induced elongated features in the larger B. fatalis. Leptodora did not induce feature elongation in the smaller B. longirostris and severely depleted this species (Chang and Hanazato 2004). Across the Baltic region of Europe, there is growing evidence that Leptodora predation may also have driven evolution of feature lengths in the larger bodied subgenus Eubosmina. Eubosmina are slightly larger in body size (total length, 300–700 mm) than Bosmina (200–500 mm) and co- occur abundantly in European lakes with Leptodora. Recent induction experiments suggest that Leptodora and Mesocyclops stimulate elongation of antennules in various Eubosmina forms (Lagergren and Stenson 2000; Kerfoot 2006). Microfossil records over thousands of years’ duration appear to show the onset of seasonal cyclomor- phosis in Eubosmina about 5000 yr ago in certain European lakes (e.g., in Grosser Ploener See and Schohsee, northern Germany; Hofmann 1978). Paleolimnological work on these lakes confirms that Leptodora was present during the cyclomorphosis interval, based on the incidence of Lepto- Fig. 8. Comparison of spine lengths (mucrones, antennules) dora mandibles and tail spines in core sediments with on individuals from laboratory common-garden clones before Eubosmina remains (Kerfoot 2006). (1986) and after (1991) Leptodora colonization. For additional examples, see Table 7. Acknowledgments Amy Hoshimoto conducted the heritability experiments at (seasonal, yearly, decadal). A seasonal component to risk Douglas Lake. Richard Kiesling sampled enclosures one summer. from invertebrate predators was reflected in significant Winfred Lampert, Director of the Max Planck Institute for month-to-month and year-to-year variability, due to the Limnology, provided support at Ploen, Germany, for the B. resident mix of invertebrate predators and fish. The longirostris comparisons. The University of Michigan Department Leptodora perturbation seemed superimposed upon a of Natural Resources caretakers at Third Sister Lake aided boat and general increase in resident invertebrate predators (Meso- enclosure studies. The authors thank two anonymous reviewers and cyclops, Chaoborus). An additional ecological cost of Luc De Meester for helpful editorial comments. This research was defense may involve adjustments to the variety of co- supported by National Science Foundation (NSF) grants BSR84- existing predators, a consideration mentioned in Tollrian 00244, BSR85-01252 (to W. C. K. and A. Kluge), Ocean Sciences and Harvell (1999). High incidence of short-featured forms (OCE) 97-12972 (joint NSF and National Oceanic Atmospheric Agency (NOAA), Keweenaw Interdisciplinary Transport Experi- (B. freyii) in 1983 coincided with low densities of ment Superior Project), OCE 97-26680 (NSF and NOAA Episodic Mesocyclops (McNaught et al. 2004) and other invertebrate Events Great Lakes Experiment Project), and DEB0083731 predators. Others have noted evidence for depressed (Biocomplexity) to WCK. Additional financial assistance was Bosmina in the presence of Epischura and Leptodora provided to ASM by the University of Michigan Rackham School (Kerfoot and Peterson 1979; Branstrator and Lehman of Graduate Studies and the Department of Biology. 418 Kerfoot and McNaught

References ———. 1978. Combat between predatory copepods and their prey: Cyclops, Epischura, and Bosmina. Limnol. Oceanogr. 23: BRANSTRATOR,D.K.,AND J. T. LEHMAN. 1991. Invertebrate 1089–1102. predation in Lake Michigan: Regulation of Bosmina long- ———. 1980. Perspectives in cyclomorphosis: Separation of irostris by Leptodora kindtii. Limnol. Oceanogr. 36: 483–495. phenotypes and genotypes, p. 470–496. In W. C. Kerfoot BRIDGEMAN, T. B., AND oTHERS. 2000. A limnological survey of [ed.], The evolution and ecology of zooplankton communities. Third Sister Lake, Michigan, with historical comparisons. University Press of New England. Lake Reservoir Manag. 16: 253–267. ———. 1987. Translocation experiments: Bosmina responses to CARTER, J. C. H., M. J. DADSWELL,J.C.ROFF, AND W. G. copepod predation. Ecology 68: 596–610. SPRULES. 1980. Distribution and zoogeography of planktonic ———. 2006. Baltic Eubosmina morphological radiation: Sensi- and dipterans in glaciated eastern North America. tivity to invertebrate predators (induction) and observations Can. J. Zool. 58: 1355–1387. on genetic differences. Archiv. Hydrobiol. 167: 147–168. CHANG, K. H., AND T. HANAZATO. 2003. Seasonal and reciprocal ———, A. S. MCNAUGHT, AND S. T. JARNAGIN. 2000. Invertebrate succession and cyclomorphosis of two Bosmina species predators and fish: Biogeographic patterns and fish dispersal (, Crustacea) coexisting in a lake: Their relation- of resting eggs. Verh. Internat. Verein. Limnol. 27: 1841– ship with invertebrate predators. J. Plankton Res. 25: 1850. 838–849. ———, G. G. MITTELBACH,N.G.HAIRSTON,JR., AND J. J. ELSER. ———, AND ———. 2004. Predation impact of Leptodora kindtii 2004. Planktonic biodiversity: Scaling up and down. Limnol. on population dynamics and morphology of Bosmina fatalis Oceanogr. 49: 1225–1228. and B. longirostris in mesocosms. J. Freshw. Biol. 49: ———, AND R. A. PASTOROK. l978. Survival versus competition: 253–264. Evolutionary compromises and diversity in the zooplankton. COUSYN, C., L. DE MEESTER,J.K.COLBOURNE,L.BRENDONCK,D. Verh. Internat. Verein. Limnol. 20: 362–374. VERSCHUREN, AND F. VOLCKAERT. 2001. Rapid, local adapta- ———, AND C. PETERSON. 1979. Ecological interactions and tion of zooplankton behavior to changes in predation evolutionary arguments: Investigations with predatory pressure in the absence of neutral genetic changes. Proc. copepods and Bosmina, p. 159–196. In U. Halbach and J. Natl. Acad. Sci. U. S. A. 98: 6256–6260. Jacobs [eds.], Population ecology. Fortschr. Zool. 25(2/3), DE MEESTER, L., A. GOMEZ,B.OKAMURA, AND K. SCHWENK. 2002. Gustav Fischer. The monopolization hypothesis and the dispersal–gene flow ———, AND L. J. WEIDER. 2004. Experimental paleoecology paradox in aquatic organisms. Acta Oecol. 23: 121–135. (resurrection ecology): Chasing Van Valen’s Red Queen DE MELO,R.,AND P. D. N. HEBERT. 1994. A taxonomic hypothesis. Limnol. Oceanogr. 49: 1300–1316. reevaluation of North American . Can. J. Zool. KORINEK, V., V. SACHEROVA, AND L. HAVEL. 1997. Subgeneric 72: 1808–1825. differences in head shield and ephippia ultrastructure within the Bosmina Baird (Crustacea, Cladocera). Hydrobio- DODSON, S. I. 1989a. Predator-induced reaction norms. Cyclic changes in shape and size can be protective. Bioscience 39: logia 360: 13–23. 447–452. LAGERGREN, R. M., M. E. HELLSTEN, AND J. A. E. STENSON. 1997. ———. 1989b. The ecological role of chemical stimuli for Increased drag, and thus lower speed: A cost for morpholog- Bosmina Eubosmina zooplankton: Predator-induced morphology in Daphnia. ical defence in ( ) (Crustacea: Cladocera). Funct. Ecol. 11: 484–488. Oecologia 78: 361–367. ———, AND J. A. E. STENSON. 2000. Chemical cues from the FALCONER, D. S. 1960. Introduction to quantitative genetics. invertebrate predator Leptodora kindtii affect the develop- Ronald Press. ment of cyclomorphotic traits in Eubosmina coregoni gibbera. FUTUYMA, D. J. 1986. Evolutionary biology. Sinauer. J. Plankton Res. 22: 1213–1219. GABRIEL, W. 1999. Evolution of reversible plastic responses: LEHMAN,J.T.,AND T. NAUMOSKI. 1986. Net community Inducible defenses and environmental tolerance, p. 286–305. production and hypolimnetic nutrient regeneration in a In R. Tollrian and C. D. Harvell [eds.], The ecology and Michigan lake. Limnol. Oceanogr. 31: 788–797. evolution of inducible defenses. Princeton Univ. Press. LEIBOLD,M.A.,AND J. NORBERG. 2004. Biodiversity in ———, B. LUTTBEG,A.SIH, AND R. TOLLRIAN. 2005. Environ- metacommunities: Plankton as complex adaptive systems? mental tolerance, heterogeneity, and the evolution of revers- Limnol. Oceanogr. 49: 1278–1289. ible plastic responses. Am. Nat. 166: 339–353. LORD, H., R. LAGERGREN,J.SVENSSON, AND N. LUNDQVIST. 2006. GRANT, J. W. G., AND I. A. E. BAYLY. 1981. Predation induction Sexual dimorphism in Bosmina: The role of morphology, of crests in morphs of the Daphnia carinata King complex. drag, and swimming. Ecology 87: 788–795. Limnol. Oceanogr. 26: 201–218. LUNTE, C. C., AND C. LUECKE. 1990. Trophic interactions of HANAZATO, T., AND M. YASUNO. 1989. Zooplankton community Leptodora in Lake Mendotae. Limnol. Oceanogr. 35: structure driven by vertebrate and invertebrate predation. 1091–1100. Oecologia 81: 450–458. LYNCH, M., AND B. WALSH. 1998. Genetics and analysis of HAVEL, J. E. 1985. Cyclomorphosis of Daphnia pulex spined quantitative traits. Sinauer. morphs. Limnol. Oceanogr. 30: 853–861. MCNAUGHT, A. S. 1993a. The role of intraguild predation, vertebrate ———, AND J. B. SHURIN. 2004. Mechanisms, effects, and scales of predation and dispersal in the organization of zooplankton dispersal in freshwater zooplankton. Limnol. Oceanogr. 49: communities. Ph.D. thesis. Univ. of Michigan, Ann Arbor. 1229–1238. ———. 1993b. Suggestions of biotic niche constraints from the HOFMANN, W. 1978. Bosmina (Eubosmina) populations of the distribution of Leptodora kindtii. Verh. Internat. Verein. Grosser Plo¨ner See and Scho¨hsee lakes during late glacial and Limnol. 25: 312–317. postglacial times. Arch. Hydrobiol. 25: 167–176. ———, R. L. KIESLING, AND A. GHADOUANI. 2004. Changes to KERFOOT, W. C. 1977. Competition in cladoceran communities: zooplankton community structure following colonization of a The cost of evolving defenses against copepod predation. small lake by Leptodora kindtii. Limnol. Oceanogr. 49: Ecology 58: 303–313. 1239–1249. Bosmina and invertebrate predators 419

PARSONS, P. A. 1973. Behavioral and ecological genetics, a study in TAYLOR, D. J., C. R. ISHIKANE, AND R. A. HANEY. 2002. The Drosophila. Clarendon Press. systematics of Holarctic bosminids and a revision that RELYEA, R. A. 2002. The many faces of predation: How induction, reconciles molecular and morphological evolution. Limnol. selection, and thinning combine to alter prey phenotypes. Oceanogr. 47: 1486–1495. Ecology 83: 1953–1964. TILMAN, D., AND P. KAREIVA. 1997. Spatial ecology: The role of RICKLEFS, R. E., AND D. SCHLUTER. 1993. Species diversity in space in population dynamics and interspecific interactions. ecological communities. Historical and geographical perspec- Princeton Univ. Press. tives. Univ. of Chicago Press. TOLLRIAN, R. 1995. Chaoborus crystallinus predation on Daphnia SAKAMOTO, M., K. CHANG, AND T. HANAZATO. 2007. Plastic pulex: Can induced morphological changes balance effects of phenotypes of antennule shape in Bosmina longirostris body size on vulnerability? Oecologia 101: 151–155. controlled by physical stimuli from predators. Limnol. ———, AND C. D. HARVELL. 1999. The ecology and evolution of Oceanogr. 52: 2072–2078. inducible defenses. Princeton Univ. Press. SHURIN, J. B., AND E. G. ALLEN. 2001. Effects of competition, WILKINSON, L. 1989. SYSTAT: The system for statistics. SYSTAT. predation, and dispersal on species richness at local and regional scales. Am. Nat. 158: 624–637. SOKAL, R. R., AND F. J. ROHLF. 1995. Biometry: The principles and practice of statistics in biological research. W. H. Associate editor: Luc de Meester Freeman. SPRULES, G. W., J. C. H. CARTER, AND C. W. RAMCHARAN. 1984. Received: 23 December 2008 Phenotypic association in the Bosminidae (Cladocera): Amended: 05 October 2009 Zoogeographic patterns. Limnol. Oceanogr. 29: 161–169. Accepted: 08 October 2009